Professional Documents
Culture Documents
By LUNA B. LEOI’OLD
T \T31,15 1
A N I ) T,~:NGT’II OF I ~ I V I C CIIANNELS
NIJICIIII~CIL I~ OF v2\itTous SIZI:S IN y’rm
UNITEI) STATICS
Ihcluding l’ributarics of Smnller Order
Mean
Draanacjr
Area, /?ZlJP?
il veraqe Total Inrliidancl Rrpiecenlatztv
I.rngth, Lenqth, Trzlmtm ies, of
order’ Nzmber Miles nllles Sq. M a . Ench Saze
l2 1,570,000 1 1,570,000 I
2 350,000 2 3 810,000 4 7
3 80,000 5 3 420,000 23
4 18,000 12 220,000 1on
5 4,200 28 11 6,000 518
6 950 64 61,000 2,460
7 200 147 30,000 11,700 Allegheny R.
8 41 388 14,000 55,600 Gila It.
9 8 777 (i,200 204,000 Columbia R.
10 1 1,800 1 ,800 1 ,250,000 Mississippi R.
The definition is that of Strahler: Ordei J is cl~nnnclwithout trihiitai-irs; order
2 is channcl with only order 1 tributnrirs, but includos only the length scgrnent be-
tween junction upstream of order l channels and junction downstrrain \\ ith another
order 2 channel.
2 The size of the order 1 rlianncl depends on sc*sleof maps used; 1hesr order niim-
hers are based on the determination of the simllcst order using maps of sc:Llc 1 62,500.
great rivers of the world. Table 2 presents a listing of the largest rivers in
each of the contincnts, though Europe is separated from Asia for purpose
of clarity. The ten rivers draining the largest areas are indicated by
numerals in the left-hand column, these numerals representing rank in
terms of drainage area. This list of ten includes five that are seldom
mentioned, the Yenisei, Ob, Lena, Amur, and Parana rivers. Four of
these are in the Soviet Union, and indeed that country has more major
rivers than the United States of America, China, and Brazil combined.
The general location of these t m largest rivers is shown on the world
map, Figure 1.
There is much onc could writc about the flow cnrricd by rivers: the
seasonal regimen, thc occurrence and movcment, of flood waves, the
TABLE 2
GREAT RIVERS OF TIIE WORLD^
Average
Rank in Drainage Discharge
Order qf Area, at Mouth,
Drainage Thousands Thousands
Area Railer Country of s q Mi of cfs
Asia
5 Ycnisei U.S.S.R. 1,000 614
ti 0b U.S.S.R. 959 44 1
7 Lena U.S.S.R. 936 547
9 Yangtze China 750 770
10 Amur U.S.S.R. 712 338
Ganges India 409 060
Bramaputma E. Pakistan 361 700
Indus W. Pakistan 358 196
Mckong Thailand 310 390
Hwang €10(Yellow) China 260 116
Kolyma U.S.S.R. 249 134
Shatt-el Arab2 Iraq 200 51
Irrawadrly Burma 166 479
Indigirlca U.S.S.R. 139 ti4
Kris hn:i India 119 69
Godavnri India 115 127
Salwecn Buriita I08 5:3
Yam U.S.S.R. 95 35
Pyaaina U.S.S.R. 74 00
Smkai (Si) China 40 127
Europe
Danube Romania 315 218
Dnieper U.S.S.R. 194 59
Dvina (Northern) U.S.S.R. 139 124
Pechora [J.S.S.R. 126 I44
Neva U.S.S.B. I09 92
Vistula Poland 76 38
Rhine Netherlands 56 75
Rhone France 37 59
Po Italy 27 51
RIVERS 515
chemistry of river waters, and a host of other subjects that make river
hydrology a valid realm for the pursuit of fundamental scientific inquiry.
I shall limit this paper to a description of the form of rivers and river
systems.
The three million miles of stream channels in the United States vary
widely in size and occur in a wide variety of topographic and geologic
circumstances. Included in this figure of total length are those high
mountain streams epitomizing wilderness beauty, -the dirty and trash-
filled channels too often coursing through our cities, and the majestic but
turbid large rivers flowing in wide valleys of central United States and
t h c flat expanses of the coastal plain.
One might suppose then, from the variety of geographic and geologic
environments through which these channels are carved, that there would
exist such a variety of river types that but little could be said about river
characteristics in general. Each geographic situation might be supposed
to have its own special kind of river and each would be a law unto itself.
Africa
2 Congo Congo 1,550 1,400
4 Nile Egypt 1 ,150 100
Zamhrzi Mozambique 500 250
Niger Nigeria 430 215
North America
3 Mississippi3 U.S.A. 1,244 611
Mackenzie Canada 697 280
St. Lawrence U.S.A. and 498 500
Canada
Nelson Canada 414 80
Yukon Canada 360 180
Columbia U.S.A. 258 256
Frazer Canada 92 113
Mobile U.S.A. 42 58
Susquehanna U.S.A. 28 38
South America
1 Amazon Brazil 2,231 4
8 Parana Argentina 890 526
Tocantins Brazil 350 360
Orinoco Venezuela 340 600
Sao Francisco Brazil 260 100
Magdalena Columbia 93 265
Uruguay Argentina and
Uruguay 90 136
Note.-This table includes only rivers that discharge over 35,000 cubic feet per
second a t the mouth. It does not include many rivers like the Colorado for example,
that drain large areas in an arid climate and therefore discharge relatively little water.
1 Rivers with average discharge of 35,000 cubic feet per second, or greater; from
As a matter of fact, thc subtle but pervasive unity that exists, despite the
tremendous diversity, is probably the most important characteristic of
river systems.
There is not only unity displayed by important similarities between
rivers in different settings, but also an amazing organization of river
systems. This in part results from a delicate balance between the forces of
erosion and the forces of resistance.
Thc characteristics of river systems which can be used t o dcmonstrate
the unity among forms fall under three gencral categories, the river
channcl, the rivcr valley, and the drainage net-the tcrm applied to the
system of branching channels ramifying to the smallest tributaries.
The first and most important aspcct of the river channel is that i t is
self-formed nrid sclf-maintained. The flowing wat cr carves the groove in
which i t flows. The watcr fashions tlic depth, the cross section, the areal
configuration, and the longitudinal profile.
Existing river channels have had a history in many respects analogous
to the history of a s p c c h in tlw biological world. A n existing river is
inherited from an cnrlicr one, which in turn was dcrived from a still
earlier progrnitor. Thc scyurncc extends backward to that time when the
contincnt:d area now being drninrd hy the river was once under the
ocean.
Thus the river channcl usually cannot tic spoken of as having becn
formcd, i n that no particular timc or place of origin of a given rivcr can be
specified. Rather, i t is best t o think that a river has evolved through
geologic timc, and that the processes of change that may be observed a t
present arc indccd the same processes by which this evolution took place.
The changes during the. river history in geologic time were brought about
by changes in the external conditions, such as a change in climate
(including glaciation), a change in elcvation of the continental mass and
the concomitant changes in vegetation, in rates and types of weathering,
and in the amounts and kinds of matcrials flowing under the influence of
gravity toward base-lcvcl-thc ocean.
The forccs involved in shaping and maintaining thc channel are related
to thc fluid flow. Thr fiow of (lie fluid cxurts :L shrar on both bed and bank
and i t is this shcnr which ~ a u s e sparticles comprising t,he bed and bank
to be moved along in trartion by the water, that! is, by pushing, rolling,
and skipping. This portion of the debris in transport is referred to as the
txdload. The same shear crratcs the turbulent cddics in the flowing fluid
which entrains some' particlcs of mattcr and throws them up into the
main mass of the fluid, thus carrying them as thc suspended load.
Thr transport of sediment drhris by rivrrs to thc oceans is a phenom-
cmon ltnown to cvcryonr. It is far lrss wcll known that the quantity of
dissolvotl mat crinls carried hy rivers t o the ocran is also very lnrgc.
13cc:iiiso t l i c transport proccss is not) visually cvidcnt, nor are dcposits of
RIVERS 51’7
dissolved materials obvious, it is hard to visualize that slightly morc tlmn
half of all the materials carried by river water from continent to occan is
carried as dissolved load. This fact was quantitatively verified in results
recently published by Durum, Heidel, and Tison, describing an inter-
national project designed to sample the dissolved load of tlie great rivcrs
of the world.
Fluid shear shapes the river channel. It can be visualized in a general
way that if a channel were vcry deep and vcry narrow there would lw
far more shearing action on the channel sides than on the channel bed.
It seems logical that, as a result, there would be a tendency in such :L
channel for erosion to take place along the banks, whereas along the hcd
tl .,re would be a tendency for insuficient shear to carry along the eroded
material. Such a channel would widen and become shallower.
An opposite extreme can be visualized, in which a channcl were very
wide and shallow. The great expanse of bed exposed to the shearing xc-
tion of the flow would tend to erode and thus degrade the bed, cstablish-
ing in time a channel of more modest breadth but greater depth. Indeed,
some such action does take place, for the width-depth ratios of rivcrs of
comparable size have a great consistency.
If the shearing force on the channel banks is sufficient to overcome the
cohesion of the bank materials, erosion takes place, and the eroded
particles are swept away from their original position and become a part
of the bed materials, there either to be moved or temporarily lodged.
On the stream bed, scour takes place when the shear exceeds some critical
value, and this occurs during relativcly high flow. At low flow i t is usual
for the shearing forces on the bed to he sufficiently small that the txd
materials do not move. Scour, then, occurs primarily during high dis-
charges. The relatively long time periods represented hy modest and low
flow are periods of but little movement of bed material.
A river increases downstream in size owing to the increase of drainage
area as tributaries enter. This increase in size is manifested not only in
the amount of water (discharge) but also by a n increase in channel
width and depth. To accommodate an increase in discharge rcsultirig
from tributary entrance, the channel may enlarge in width or in depth,
or the increased flow could he achicvcd by an increase in flow velocity.
Because discharge (cubic feet per second) is the product of cross-sectional
area times velocity
Q = av
f
W
n
@-v-v-
Loup (Nebr )
Mississippi. Main Stem
Madras Irrigation Canals (India)
FIG.2. Width, dcptli, and vclocity in rclation to incan annual discharge, as discharge
increascs downstream in various river systems.
velocity. The parallelism of lines in each of the plots shows that the rates
of inrrcasc with discharge are similar among the various rivers.
Width increases downstream faster than depth, and depth faster
than velocity, the latter being very nearly constant along the river
lcngth. The increase in width-to-depth ratio can be sccn in Ipigure 3, in
which cross sections of five channels arc drawn with no cxaggeration in
RIVERS 519
vertical scale but a t different scales so that the widths appear equal on
the printed page. In actuality the creek represent,cd in the upper sketch
has a width of only 24 feet whereas thc river in the lower sketch is a large
river having a width of 1153 feet.
It had long been supposed that the speed of water in a river decreased
downstream. This supposition was based on the fact that river slope
decreases downstream and, generally speaking, the size of bed material
also decreases from boulders or cobbles in the headwaters to sand, silt, or
clay in the downstream portions of large rivers. It 1i:d long hecn argued,
without reference to measurement data, that the supposed downstream
decrease in water velocity was causally related to the inahilit,y of a river
to carry coarsc' material a t the flat gradient, cxisting in downstream
-
0 2 4 FEET
0 10 20 3 0 FEET
u
-- Rock River, Wis.
W = 2 8 0 Feet 0 20 40 FEET
u
Potornac River, Md.
\ c
W=1153 Feet 0 100 200 FEET
W
FIG.3. Cross sections of five rivers of different size, no vertical csitggerntion, but
scaled so that width is cqual on the printed page.
reaches. Mountain creeks appear lo flow mom swiftly than the deep
water of large rivers, but the appearance of flowing water is a test rather
of kineticity than of flow speed. When my collcngucs and I showed by
measurement data that, in fact, velocity tends to incrc:w or remain
constant downstream in most rivers, a new rationale was needed t o
explain thc observed relationships.
It now appears that, the constancy or slight increase of vclocity down-
stream results from the fact that river depth increases fastcr than river
slope decreases. But this still does not, explain liom the requisite changes
are divided among the pertinent parameters, nor why this division is so
520 AMERICAN SCIJSNTIST
10 -
w
F
W
LL
5-
z
+-
I
I?
W
r
w 0 -
c7
a
u
-5 -
FIG.4. Scour folloxed by fill during the passage of flood during snon iiielt season, Colorado River at Lees Ferry, Arizona, 1956.
522 AMEIEICAN scrmmsr
Bagnold's 1956 research has shown, first theoretically and then by
experiment, that grains in traction create a downward-acting stress which
actually tends to hold the exposed grains on the stream bed and thus
prevent their erosion. This downward force results from grain collision
during transport. Collisions between grains tend t o knock an individual
grain in a direction that may be forward, up, or down. There will be
enough grains directed downward against the bed actually to create a
downward stress which tends to prevent further erosion of bed grains.
Until Bagnold's work, i t had not been recognized that hydraulic
relations developed for clear water would not apply directly to water
carrying a debris load. The forces within a fluid are altered by the trans-
port of sediment debris.
Knowledge of sediment transport and its relation to river hydraulics
has been slow to develop, both in theory and in practice, owing t,o the
fact that i t is still impossible to make consistent and accurate measure-
ments of the amounts of debris being transportcd as bedload in natural
rivers. Instruments have long been in use that measured adequately the
sediment being transportcd in suspension, but bedload, bein,0' concen-
trated in a thin zone near the stream bottom, is interrupted by the
insertion of any obstacle, no matter how streamlined. The inscrtiori of a
sampling device a t the stream bed changes the conditions of transport
locally in such a manner that a representative sample of the bedload in
transport is impossible. Though rates of bcdload transport can be suc-
cessfully measured in experimental flumes, the conditions represented in
the laboratory are much more limited than those extant in the field.
Water discharge in a river fluctuates widely from year to year, season
to season, and day to day. Thus the forces governing erosion and depo-
sition are highly variable. The channel flows at shallow depth most of the
time and is filled to the top of the banks relatively scldom. On still less
frequent occasions, the discharge exceeds the channel capacity and flows
overbank. An overbank flow is a flood by definition. Reccntly it was
shown that the frequency of the bankfull condition is surprisingly uni-
form among rivers in diverse settings and among rivers of widely different
size. The bankfull stage may he expeibed on the average about once a
ycar to once in two years.
7'0 state this in another wily so that the prxticnl application is ap-
parent, the river channel is largc cnough to :wcommod:ite a11 the water
coming from the drainage area only in thr rclativcly frequent, w e n t . The
flat area bordering most channcls-t,bc flood pl:iirl-must flood to some
extent on the average every other year. To ovcrflom the flood plain is an
inherent charactcristjic of u river.
It has Iwcn surmised that vhanncl formation and maintenance are
accomplished primarily during periods of flood flow. This would seem to
follow from the general observation that many of the most obvious
I
EXPLANATION
0
0 98 ,
‘In
- -
1
1 10 100 1000 5 10 100 1000 10,000 1c
C H A N N E L WIDTH, IN FEET MEAN RADIUS OF CURVATURE, IN FEET
A B
Fig. 5. Relations between meander length and channel width (left diagram) and mean radius of curvature (right diagram)
524 AMEKICAN scIEwrisr
changes in form or configuration are accomplished by the catastrophic
event-by the flood of large magnitude. To the extent this were true, it
would be inferred that the events most worthy of study in attempting
to isolate the general controlling principles would be extraordinary
occurrences.
As recently as 19G0, Wolman and Miller introduced another conception
that has had an impact on thinking about rivers and, indeed, alters the
view of the action of geomorphic processes in general. The principal
work in fashioning landscape forms is not done during the frequent event
or the predominant condition because the processes acting are incompe-
tent to have any significant effect. Nor is the principal work accomplished
by the very effective but very rare event-the catastrophic incident.
Rather, i t appears that the landscape is fashioned primarily by the
cvent of intermediate freyucncy and effectiveness.
For example, the total sediment load transported by a river past a
given point over a period of time is the sum of the amount transported
during a variety of flow events; that is, during many days of low flow, a
modest number of days of intermediate flows, and a decreasing number
of increasingly great floods. The data indicate that the intermediate
flows have a cumulative significance far greater than those of either
greater or smaller magnitude, 1 ecause the product of effectiveness times
frequency is grcatest. This concept is a fundamental contribution to
thinking about landscape moulding processes.
Thcre arc other attributes of river channels which also demonstrate the
unity that exists among channels, even channels widely different in size.
Nearly all natural channels are sinuous to some extent. Indeed, natural
channels w e rarely straight for a distance greater than ten channel
widths. Not only do channels exhibit a more or less regular aspect of
sinuosity, hut also the size of the curves assumed by a channel bear a
constant rclationship to the channel itself. Small channels wind in small
curves and large channels in large curves.
Measurement data show that there is quite a constant relationship
between channel width and the radius of the curves which the channel
exhibits. Also, the meander length or wave length is generally propor-
tional to channel width. These characteristics can be seen on Figure 5 ,
where the meander length is plotted separately against channel width
and against the mean radius of curvat,ure. It can be seen that there is
essentially a linear relationship between the wave length of the channel
curves, channel width, and the radius of curvature. This linear relation
occurs not only in most natural rivers but also to meltwater channels
developed on the surface of a glacier and in the flow path of the gulf
stream which is not confined by channel banks.
The consistency of the relation of radius of curvature to channel
width is indicated by the following measurement data. I n a sltmple of 50
525
- 1
FLOOD PLAIN
Approximate elev 9 2 feet
\ Ji
8 feet
Average width r-;i
/
Approximate elev 94 0 feet
0 20 FTET
/I
rivers of various sizes the median value of the ratio of curvature divided
by width was 2.7, the mean 3.1, and two-thirds of the cases occurred
between the values of 1.5 and 4.3. The tendency for a constant ratio of
radius to width makes all rivers look quite similar on planimetric
maps, as is suggested by Figure 6. I n fact, when one inspects a plani-
metric map of a river without first glancing a t the map scale, i t is not
immediately obvious whether the river is large or small owing to this
tendency for a similar ratio of radius to width.
When the sinuous bends of a river are relatively symmetrical the
channel is said to be meandering. Such a large variation in degree of
sinuosity exists in natural channels that the definition of a meandering
stream-in contrast with one which is merely sinuous-becomes a
matter of arbitrary definition. There is represented in nature, then, an
unbroken continuum between slightly wavy channels and well developed
meanders which can assume the form of horseshoe loops. Owing t o the
fact that the meandering character of river channels has long been
considered to be associated in basic mechanics with the movement of
sediment, the observation that sediment-free channels carved by melt-
water on the surface of a glacier can also meander is of considerable
interest. Tentatively, the mechanical principles resulting in the meander-
ing form of meltwater channels in the absence of sediment load and
those causing the meandcring of rivers are probably the same. The
similarity in channel form implies that the basic cause is essentially
hydrodynamic, rather than a direct result of bank erosion and the lateral
transfer of sediment across the channel.
The mechanics of a meandering channel are by no means well under-
stood. Certain physical characteristics have been known for a long time
but the basic mechanism is still elusive. Measurements made in meander-
ing streams and in curved flumes allow the construction of a generalized
picture of the flow pattern in a meander which is presented in simplified
form in Figure 7. The isometric view of the two principal components of
velocity a t various positions in the bend show the main features. The
scale of the sketch is such that the super-elevation of the water surface in
the bend does not show in the diagranils but is implied in the velocity
distribution.
The velocity in a meander crossover, or point of inflection, is not
symmetrically distributed. As would be expected, proceeding down-
stream from the axis of the bend the thread of maximum velocity is much
closer to the concave bank than to the center of the channel. The high
velocity, moreover, continues to hug tthis side through the point of in-
flection of the curve. At least in narrow channels, a cross-channel velocity
component is directed toward thc convcx bank (usually called the point
bar) near the bed and toward the concave bank near the surface.
Continuity requires, then, that surface watcr plunge downward near
HIVEItS 527
b 5
Generalized velocity
distribution
Generalized surface
streamlines
the concave hank and that some bed water emerge at the surface near t
Convex bank. This circwlatory motion in the cross-sectional plane of
channel mas first observed and cxplaincd by Thomson in 1879. It resul
from the largclr centrifugal force on fast-moving siirface parcels than
dower-moving ones ncar the bed. The motion gives to an individual
mater parcel a pat,h resembling a helix. As a result of this circulatory
motion, material a t the bed tends t o be swept toward the convex bank,
thus to be deposited in part on the point bar.
The accretion of matcrial on tkc point bar gradually pushes the convex
hank into the channel and this tendency is compensated by a tendency
for erosion of the concave hank. Thus, it is usual for a river channel
gradually to migrate latcrally across thc valley floor. During such lat
migration the channel width remains thc same. An example of the b
ing of the point, bar balanced by bank erosion on the opposite side o
stream is demonstrated by a series of observations made on Watts
13ranch, a small stream near Washington, D.C., for which the cross
sections during a period of nine years arc presented in Figure 8.
The manner in which a channel moves across thc valley floor, eroding
one hank and building a nearly flat flood plain on the other, all the while
maintaining a cross scction similar in shape and size, is another aspcct of
t,hc dynamic cquilibrium that apprars to charactcrizc many channel
systems.
Another aspect of the unil,y rxhihitcd by stream systems is cxemplifi
in the branching network of tributaries comprising what is called t
drainage net. I n 1945, Robert E. Horton showed that thc size of a giv
stream is related to the number of streams of that size, t h
length, and average slope. There is a very constant quantitative rela
tionship between these factors which can best be dernonstratcd by so
actual data. Figure 0 is a sketch to define stream order, a measure
stream size. Horton defined order in the following manner. The small
unbranched tributary in the headwaters is, by convention, designat
order 1 ; a stream which has only tributaries of order 1 size is called scco
order; a third-order stream is one in which all tributaries arc of second
order or smaller, ctc. Figure 10 shows that there is a logarithmic relation
hctween stream order and length of streams of a given order and be-
tween stream order and numbers of streams. After this logarithmic
relation within the drainage network was pointed out, by Florton, it has
heen testcd by many workers and fouiid applicable, essentially without
cxccption, to drainage basins of all sizes, including t,he very smallest rills
and large river basins.
One of the basic problems in the study of rivers is the identification of a
general lam or princaiplc, which provides a rationale in physical terms
for the different types of unity exhibited by river channcls and the
nctworks which they comprise. It is only too easy to speak of the rcsiilts of
>
>
U
7
3
7
P
0
3
U
*
3
8
m
N
W
N
d
N
N
hl
0
N
L
530 AMXl1IChN SCIICNTIST
read in the popular press, as a result, Ihat the river W Con~aI rampage,
overflowing its banks and wrcakirlg havoc. on man and his strnctures.
ItIVEIlS 531
express portions or the totality of the actions of the system which appear
to have definite tendencies toward a direction of development, or ten-
dencies which tend to maintain the average conditions observed.
One way in which this has been approached has been a description of
thc river system as having developmental stages through time. This
was one of the basic concepts introduced by William Morris Davis, who
described the geographical cycle as consisting of three principal stages
comparable to the stages in the life of man. The stage of youth was
EXPLANATION
0 Average values, Arroyo Caliente I
0-
Arroyo de 10s Frijoles
V Arroyo de las Trampas to Rio Santa Fe
0 Rio Santa Fe. Rio Galisteo, and
Caliada Ancha (del oriente)
6 7 8 9 1 0 I1 1 2 3 4 5 6 7 8 9 1 0 1 1
STREAM ORDER
Fro. 10. Relation of strcitiri lcrigth arid riuinhcr of streams to stream order in bitsins of
1 I th order near Santa Fc, New Mexico.
during this relatively long period in the cycle the rivcr was considcred
to be unable to do large amounts of work in erosion as a result of rela-
tively low relief and flat gradients. Similarly, a t any one point in time the
headwaters of a river system could be in the stage of youth, the central
portion in the stage of maturity, and the downstream portion in the
stage of old age. This characterization of a river system was extremely
useful for teaching purposes, but from a quantitative standpoint many
aspects are difficult to particularize.
There followed, then, more detailed consideration of the concept of
equilibrium, or grade, which Davis had thought of as characteristic of
the stage of maturity. A graded stream was described by J. Hoovcr
Mackin as one “in which, over a period of years, slope is delicately
adjusted to provide, with available dischsrge and with prevailing channel
characteristics, just the velocity required for the transportation of the
load supplied from the drainage basin. The graded stream is a system in
equilibrium; its diagnostic characteristic is that any change in any of the
controlling factors will cause a displacement of the equilibrium in a
direction that will tend to absorb the effect of the change ( I 948, p. 471) .”
Wolman extended this concept and pointcd out that the concept of
grade had two aspects, involving adjustability and stability. I n one
respect, the quasi-equilibrium of a graded stream is, as RIackin pointcd
out, characterized primarily by its ability to adjust to changes in external
conditions. Wolman believed that a concept of quasi-equilibrium may be
characterized by the stability of the system, but that adjustability
perhaps was no less important a criterion.
Another aspect which, under certain circumstances, had been used to
describe the underlying tendency for a river toward a quasi-equilibrium
condition was what might be called the principle of least work. There are
certain characteristics, particularly hydraulic ones, which point to thc
tendency for minimization of work in a system.
Recently, Lnngbein and I considered the various factors governing
the direction of river work and reasoned as follows. Inspecting the
physical principles which recur most often in analyses of differcnt aspects
of rivcr morphology, the basic equatims which are most promincnt arc
the equation of conservation and the equation of continuity. These laws
are so ohviously general that they characterize each element or rcach in
any fluvial system. They must also characterize each unit of any path
and a t each instant in time. However necessary these equations might
be, they arc insufficient to explain the paths of particles moving in the
river systems, or the relation between one part of the path t o another.
Thercforc, they can alone tell us nothing about the surface form of the
landscape, nor can they treat cmmpletely the progressive development or
change of form with time.
There must, thercforc, be some other law of cqual gcncrality which
IZIVERS 5:3:;
would deal with energy distributions and their relation l o changes of land
forms in space and time. We proposed that implications of the stwind
law of thermodynamics satisfy thesc requiremmts.
The dcvelopment of landscape involves not only the total available
cnergy hiit also its distribution. The lattcr may lie described by the
caoncept of entropy, adapting that term from the comparable concept in
thermodynamics. The essential idea is t h t the entropy of the system
is a function of the distribution or availability of energy within the sys-
tcm and not a function of the total energy within it. Thus, entropy has
come to concern order and disorder, an aspcct of the conception which
has been utilized in physical chemistry and information theory. The
degree of order or disorder in a system may be described in terms of the
probability or improbability of any observed state.
An increase in entropy is, then, a measure of the decrease in avail-
ability of the energy in the sense that a certain amount of energy is no
longer available for conversion to mechanical work. The distribution of
energy may be stated in terms of the probability of the given distribution
occurring relative to alternative distributions possible.
The statistical conception of entropy is that aspect which appears to
have application to geomorphic systems. The distribution of energy in a
geomorphic system is one way of expressing the relative elevation of
particles of water and of sediment which gradually will, in the process of
landscape evolution, move downhill toward base level. The longitudinal
profilc of the river, for example, is a statement of the spatial distribution
of stream-bed materials with regard to their elevation and, thus, with
regard to their potential energy.
I n thermodynamics, hcat energy is referred to absolute! temperature
as a base. The absolute temperature defines an absolute limit or base
datum, the situation in which molecular motion becomes zero. It is,
then, the base level, or the datum, against which the energy content of a
thermal system can be measured. Systems in geomorphology also have a
base datum with regard to the distribution of energy. This base datum
is the datum of elevation, in most cases represented by mean sea levcl.
Rut the classical treatment of entropy in thermodynamics deals with
closed systems in which cntropy continuously increases to a maximum
stationary level at equilibrium. I n closed systems there is no loss or
addition of energy.
Geomorphic processes operate, on the other hand, in open systems
in which energy is being added in some places while in other places
energy is being degraded to heat and is thus lost insofar as further me-
chanical work is concerned. A river system, then, we consider to be an
example of a n open system, defining the system as the water and the
debris in the river channel. As the water flows down the channel it gives
up potential energy which is converted first to kinetic energy of the
554 AMRILICAN SCIENTIST
flowing water and which in the process of flow is dissipated into heat
along the channel margins. Precipitation brings increments of energy into
the system because water enters atj various elcvations and thus with
various amounts of potential energy. Heat is lost by yonvcrtion, conduc-
tion, or radiation, yet the channel may be considcrrd in dynamic
equilibrium.
The steady state possible in open systems differs from t]be steady state
of static equilibrium of closed systems. Wc elqimte, therefore, the term
steady state with dynamic equilibrium in geomorphology as defined
early in the century by Gilbert and, more recently, by 1I:wk. I n :in open
system in dynamic eqiiilibrium the rate of incrmse of entropy in the
system is zero. The continuity of mtropy then takes t,lie form that the
rate of outflow of entropy equals the rate of intrrnd generation of
entropy. I n a river system in equilibrium, the rate of outflow of entropy
is the rate of dissipation of energy as heat. This is equated t o the rate of
generation of entropy represented by t,hc energy gradient toward h s e
level. The stlationmy state of an open system is charac
conditions in which the rate of production of entropy per unit volume
of flow is a minimum compatible with the conditions imposed on the
system. Hence a stable system corresponds to one of least work, a con-
ception we were able to demonstrat,c in mathematic*al terms for certain
conceptual models.
The argument was developed that, witth rcyx to the longitudinal
profile of the river, the probability of the particular combination of
values of energy in various unit distances along the course of the open
system of the river could be described as proportional to the negative
exponential of its ratio t o the total energy of all possible st,ntcs. This
led to the conclusion that the most probable scquence of energy losses in
successive units of river lengths corresponds to a uniform increase in
ent,ropy in each unit, length along the river system. When this spccifica-
tion is fiilfilled without constraintj on river length tbc 1ongitndin:rl profile
of the river tends to become exponential in form, :I rcsiilt i n :rgrerment)
with many actual river profiIes.
I n the development of this probabilist,ic argummt,, random-walk
models were constructed to demonstrate several aspects of the river
system. Longitudinal profiles romparablc to thosc of rivers may hc
constructed hy random walks. As one example, beginning at an arbitrary
point on Cartesian graph paper, the walk inovcs from t,he initial point in
unit steps to the right wit,h various probabilities of a downward move-
ment simultaneously. Such random walks may be constructed from a
series of random numbers or with special packs of' cards, and the results
from such mechanical construction of random walks wert caompared
with a larger number of random walks developed on a digital computer.
An example is shown in Figure 11. In this model an initial point was
ItIV IC 11s 535
a t an clcvation of 5 units above the h s c lcvel. I n successive moves,
lhc walk proceeds to the right on(’ m i t at :L time. There are G levels
(0, 1,2,3, 4, 5 ) . Thc deck of cards uscd to draw thc random numbers was
so :Lrr:tngcd t,hat the prohntdil y of :L down\vard step during a unit step
to the right WLH U / G . Thus thc prol)aldit y of n don.nward stcp decreases
\vi t h clevat,ion.
On Figure 11 two sample random walks are shown marked by crosses
and cirrles respectively. The average vertical position a t any abscissa
v:ilire was determined by averaging values for a large number of in-
dividual walks, and the average values define the profilc drawn in a heavy
line. This average proiilr is exponential in form, :L general form exhibited
by many natural stream profiles.
Certain ch:Lrnctcristics of drainage iwt works wwe also exemplified by
rnndom walks in varioirs typcs of models. lcigure 12 shows a random-
2-7
-7
o e e
I-
5 2
2
W
-I
w
1
0
0 5 10 15 20
DISTANCE, X
. I 1 . S:mlplc r:intlom w:ilks used in thc grneration of nn nvcr:~ge11~11git~idir~:d
FIG. profilc
of :t rivrr.